Graphs provide essential means for organizing and analyzing complex equipment data. Although link prediction techniques have been widely applied to enhance knowledge graphs, existing methods still show room for improvement in accuracy, especially when dealing with sparse data. To address this, we introduce ELPGPT (Large Language Models Enhancing Link Prediction in Electrical Equipment Knowledge Graph), a novel approach that integrates large language models into link prediction to enhance the accuracy of relation prediction within electrical equipment knowledge graphs. The core of the ELPGPT method lies in the combination of large language models with traditional knowledge graph link prediction techniques. By leveraging the deep semantic understanding capabilities of large language models, this method effectively extracts relational features and enhances the handling of sparse data. Additionally, we employ a Retrieval-Augmented Generation (RAG) approach, which, by integrating external data sources, further enhances the precision and relevance of predictions. Experiments on the Electrical Equipment Knowledge Graph (EEKG) demonstrate that ELPGPT significantly improves performance across several metrics, including Hit@k, Mean Rank (MR), and Mean Reciprocal Rank (MRR). These results validate the effectiveness and potential applications of this method in the domain of link prediction for electrical equipment knowledge graphs.
Random walk-based link prediction algorithms have achieved desirable results for complex network mining, but in these algorithms, the transition probability of particles usually only considers node degrees, resulting in particles being able to randomly select adjacent nodes for random walks in an equal probability manner, to solve this problem, the asymmetric influence-based superposed random walk link prediction algorithm is proposed in this paper. This algorithm encourages particles to choose the next node at each step of the random walk process based on the asymmetric influence between nodes. To this end, we fully consider the topological information around each node and propose the asymmetric influence between nodes. Then, an adjustable parameter is applied to normalize the degree of nodes and the asymmetric influence between nodes into transition probability. Based on this, the proposed new transition probability is applied to superposed random walk process to measure the similarity between all nodes in the network. Empirical experiments are conducted on 16 real-world network datasets such as social network, ecology network, and animal network. The experimental results show that the proposed algorithm has high prediction accuracy in most network, compared with 10 benchmark indices.
This paper presents a novel link prediction approach, termed Basic-Structural Similarity Link Prediction (BSSLP), designed to address the zero-similarity problem in complex networks. BSSLP integrates basic similarity, which establishes a nonzero baseline for all node pairs, with structural similarity that captures both local and intermediate topological features. This integration effectively mitigates challenges such as cold start and sparse network prediction. Through extensive experiments on nine real-world networks, BSSLP consistently outperforms seven benchmark methods, achieving an average AUC improvement of 5.17%. The method demonstrates robust performance across various network structures and maintains high prediction accuracy under different training set proportions. By providing nonzero similarity estimates for all potential edges, BSSLP significantly enhances the prediction of new connections and offers deeper insights into network dynamics.
Link prediction aims to identify missing links within static networks or estimate the probability of emerging links in dynamic networks, representing a critical and challenging research direction in complex network. Many similarity-based methods have been established from various viewpoints, however, there are relatively few methods that take into account both the path information between node pairs and the nodes along the path. To fill this gap, we propose two novel link prediction algorithms, namely LPRA and LPH. The core idea is that the similarity between node pairs is closely related to the local paths connecting the two nodes and the resource transition capability of the nodes along those paths. The algorithms utilize the local paths with adjustable lengths between node pairs and the topological information of the nodes on the path to calculate the similarity index, which incorporate the resource transition capabilities of all the nodes on the possible paths between node pairs. We conducted multiple groups of comparative experiments on 10 real-world networks to validate the effectiveness of the proposed algorithm. Experimental results demonstrate that LPRA and LPH outperform nine classical methods and five recently popular methods. Moreover, LPRA demonstrates a significant difference level (P-value <0.05) compared to most of the methods in the ANOVA of AUC accuracy, indicating a statistically significant performance advantage of LPRA.
Predicting missing links in complex networks is of great significance from both theoretical and practical point of view, which not only helps us understand the evolution of real systems but also relates to many applications in social, biological and online systems. In this paper, we study the features of different simple link prediction methods, revealing that they may lead to the distortion of networks’ structural and dynamical properties. Moreover, we find that high prediction accuracy is not definitely corresponding to a high performance in preserving the network properties when using link prediction methods to reconstruct networks. Our work highlights the importance of considering the feedback effect of the link prediction methods on network properties when designing the algorithms.
Link prediction measures have been attracted particular attention in the field of mathematical physics. In this paper, we consider the different effects of neighbors in link prediction and focus on four different situations: only consider the individual’s own effects; consider the effects of individual, neighbors and neighbors’ neighbors; consider the effects of individual, neighbors, neighbors’ neighbors, neighbors’ neighbors’ neighbors and neighbors’ neighbors’ neighbors’ neighbors; consider the whole network participants’ effects. Then, according to the four situations, we present our link prediction models which also take the effects of social characteristics into consideration. An artificial network is adopted to illustrate the parameter estimation based on logistic regression. Furthermore, we compare our methods with the some other link prediction methods (LPMs) to examine the validity of our proposed model in online social networks. The results show the superior of our proposed link prediction methods compared with others. In the application part, our models are applied to study the social network evolution and used to recommend friends and cooperators in social networks.
As a significant problem in complex networks, link prediction aims to find the missing and future links between two unconnected nodes by estimating the existence likelihood of potential links. It plays an important role in understanding the evolution mechanism of networks and has broad applications in practice. In order to improve prediction performance, a variety of structural similarity-based methods that rely on different topological features have been put forward. As one topological feature, the path information between node pairs is utilized to calculate the node similarity. However, many path-dependent methods neglect the different contributions of paths for a pair of nodes. In this paper, a local weighted path (LWP) index is proposed to differentiate the contributions between paths. The LWP index considers the effect of the link degrees of intermediate links and the connectivity influence of intermediate nodes on paths to quantify the path weight in the prediction procedure. The experimental results on 12 real-world networks show that the LWP index outperforms other seven prediction baselines.
Link prediction in social networks has become a growing concern among researchers. In this paper, the clustering method was used to exploit the grouping tendency of nodes, and a clustering index (CI) was proposed to predict potential links with characteristics of scientific cooperation network taken into consideration. Results showed that CI performed better than the traditional indices for scientific coauthorship networks by compensating for their disadvantages. Compared with traditional algorithms, this method for a specific type of network can better reflect the features of the network and achieve more accurate predictions.
In complex networks, the existing link prediction methods primarily focus on the internal structural information derived from single-layer networks. However, the role of interlayer information is hardly recognized in multiplex networks, which provide more diverse structural features than single-layer networks. Actually, the structural properties and functions of one layer can affect that of other layers in multiplex networks. In this paper, the effect of interlayer structural properties on the link prediction performance is investigated in multiplex networks. By utilizing the intralayer and interlayer information, we propose a novel “Node Similarity Index” based on “Layer Relevance” (NSILR) of multiplex network for link prediction. The performance of NSILR index is validated on each layer of seven multiplex networks in real-world systems. Experimental results show that the NSILR index can significantly improve the prediction performance compared with the traditional methods, which only consider the intralayer information. Furthermore, the more relevant the layers are, the higher the performance is enhanced.
Traditional link prediction indices focus on the degree of the common neighbor and consider that the common neighbor with large degree contributes less to the similarity of two unconnected endpoints. Therefore, some of the local information-based methods only restrain the common neighbor with large degree for avoiding the influence dissipation. We find, however, if the large degree common neighbor connects with two unconnected endpoints through multiple paths simultaneously, these paths actually serve as transmission influences instead of dissipation. We regard these paths as the tie connection strength (TCS) of the common neighbor, and larger TCS can promote two unconnected endpoints to link with each other. Meanwhile, we notice that the similarity of node-pairs also relates to the network topology structure. Thus, in order to study the influences of TCS and the network structure on similarity, we introduce a free parameter and propose a novel link prediction method based on the TCS of the common neighbor. The experiment results on 12 real networks suggest that the proposed TCS index can improve the accuracy of link prediction.
Urban road network (referred to as the road network) is a complex and highly sparse network. Link prediction of the urban road network can reasonably predict urban structural changes and assist urban designers in decision-making. In this paper, a new link prediction model ASFC is proposed for the characteristics of the road network. The model first performs network embedding on the road network through road2vec algorithm, and then organically combines the subgraph pattern with the network embedding results and the Katz index together, and then we construct the all-order subgraph feature that includes low-order, medium-order and high-order subgraph features and finally to train the logistic regression classification model for road network link prediction. The experiment compares the performance of the ASFC model and other link prediction models in different countries and different types of urban road networks and the influence of changes in model parameters on prediction accuracy. The results show that ASFC performs well in terms of prediction accuracy and stability.
As an elementary task in statistical physics and network science, link prediction has attracted great attention of researchers from many fields. While numerous similarity-based indices have been designed for undirected networks, link prediction in directed networks has not been thoroughly studied yet. Among several representative works, motif predictors such as “feed-forward-loop” and Bi-fan predictor perform well in both accuracy and efficiency. Nevertheless, they fail to explicitly explain the linkage motivation of nodes, nor do they consider the unequal contributions of different neighbors between node pairs. In this paper, motivated by the investment theory in economics, we propose a universal and explicable model to quantify the contributions of nodes on driving link formation. Based on the analysis on two typical investment relationships, namely “follow-up” and “co-follow”, an investment-profit index is designed for link prediction in directed networks. Empirical studies on 12 static networks and four temporal networks show that the proposed method outperforms eight mainstream baselines under three standard metrics. As a quasi-local index, it is also suitable for large-scale networks.
Link prediction based on node similarity has become one of the most effective prediction methods for complex network. When calculating the similarity between two unconnected endpoints in link prediction, most scholars evaluate the influence of endpoint based on the node degree. However, this method ignores the difference in contribution of neighbor (NC) nodes for endpoint. Through abundant investigations and analyses, the paper quantifies the NC nodes to endpoint, and conceives NC Index to evaluate the endpoint influence accurately. Extensive experiments on 12 real datasets indicate that our proposed algorithm can increase the accuracy of link prediction significantly and show an obvious advantage over traditional algorithms.
Link prediction has been widely applied in social network analysis. Existing studies on link prediction assume the network to be undirected, while most realistic social networks are directed. In this paper, we design a simple but effective method of link prediction in directed social networks based on common interest and local community. The proposed method quantifies the contributions of neighbors with analysis on the information exchange process among nodes. It captures both the essential motivation of link formation and the effect of local community in social networks. We validate the effectiveness of our method with comparative experiments on nine realistic networks. Empirical studies show that the proposed method is able to achieve better prediction performance under three standard evaluation metrics, with great robustness on the size of training set.
Link prediction is an important issue for network evolution. For many real networks, future link prediction is the key to network development. Experience shows that improving reliability is an important trend of network evolution. Therefore, we consider it from a new perspective and propose a method for predicting new links of evolution networks. The proposed network reliability growth (NRG) model comprehensively considers the factors related to network structure, including the degree, neighbor nodes and distance. Our aim is to improve the reliability in link prediction. In experiments, we apply China high-speed railway network, China highway network and scale-free networks as examples. The results show that the proposed method has better prediction performance for different evaluation indexes. Compared with the other methods, such as CN, RA, PA, ACT, CT and NN, the proposed method has large growth rate and makes the reliability reach the maximum at first which save network construction resources, cost and improve efficiency. The proposed method tends to develop the network towards homogeneous network. In real networks, this structure with stronger stability is the goal of network construction. Therefore, our method is the best to improve network reliability quickly and effectively.
As a research hotspot of complex network analysis, link prediction has received growing attention from various disciplines. Link prediction intends to determine the connecting probability of latent links based on the observed structure information. To this end, a host of similarity-based and learning-based link prediction methods have been proposed. To attain stable prediction performance on diverse networks, this paper proposes a supervised similarity-based method, which absorbs the advantages of both kinds of link prediction methods. In the proposed method, to capture the characteristics of a node pair, a collection of structural features is extracted from the network to represent the node pair as a vector. Then, the positive and negative k-nearest neighbors are searched from existing and nonexisting links, respectively. The connection likelihood of a node pair is measured according to its distances to the local mean vectors of positive and negative k-nearest neighbors. The prediction performance of the proposed method is experimentally evaluated on 10 benchmark networks. The results show that the proposed method is superior to the compared methods in terms of accuracy and stableness.
Link prediction in temporal networks has always been a hot topic in both statistical physics and network science. Most existing works fail to consider the inner relationship between nodes, leading to poor prediction accuracy. Even though a wide range of realistic networks are temporal ones, few existing works investigated the properties of realistic and temporal networks. In this paper, we address the problem of abstracting individual attributes and propose a adaptive link prediction method for temporal networks based on H-index to predict future links. The matching degree of nodes is first defined considering both the native influence and the secondary influence of local structure. Then a similarity index is designed using a decaying parameter to punish the snapshots with their occurring time. Experimental results on five realistic temporal networks observing consistent gains of 2–9% AUC in comparison to the best baseline in four networks show that our proposed method outperforms several benchmarks under two standard evaluation metrics: AUC and Ranking score. We also investigate the influence of the free parameter and the definition of matching degree on the prediction performance.
A multilayer network is a useful representation for real-world complex systems in which multiple types of connections are formed between entities. Connections of the same type form a specific layer of the network. We propose a novel framework for predicting links in a target layer of a multilayer network by taking into account the interlayer structural information. The method depends on the intuitive assumption that two node pairs in the target layer tend to have similar connection patterns if these pairs of nodes are similar. Further, the prediction accuracy will be improved in the target layer if the structural information of the copies of the node pairs in relevant layers is employed. We demonstrate the effectiveness of the proposed method experimentally by applying it to both simulated and real-world multilayer networks.
Link prediction, aiming to find missing links in a current network or to predict some possible new links in a future network, is a challenging problem in complex networks. Many existing link prediction algorithms perform the task by optimizing the node similarity measures, and then determining the possibility of the link between any pair of similar nodes. In this paper, we propose a novel node similarity index named heterogeneous degree penalization (HDP), which incorporates the quasi-local structure information of extending neighborhood of each pair of nodes to be predicted and the clustering coefficient of their common neighbors. For specific networks with different statistical properties, we can achieve a good performance of link prediction through adjusting the penalty weights. The experiment results show that, comparing with the other existing approaches, the proposed method can remarkably improve the accuracy of link prediction.
Link prediction is a fundamental study with a variety of applications in complex network, which has attracted increased attention. Link prediction often can be used to recommend new friends in social networks, as well as recommend new products based on earlier shopping records in recommender systems, which brings considerable benefits for companies. In this work, we propose a new link prediction algorithm Local Neighbor Gravity Model (LNGM) algorithm, which is based on gravity and neighbors (1-hop and 2-hop), to suggest the formation of new links in complex networks. Extensive experiments on nine real-world datasets validate the superiority of LNGM on eight different benchmark algorithms. The results further validate the improved performance of LNGM.
Please login to be able to save your searches and receive alerts for new content matching your search criteria.